EFAS replication

[usereight8]

I am trying to replicate EFAS using first the following cold run: https://github.com/usereight8/lisflood/blob/main/cold.xml

I am using the static maps and meteorological forcings from: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/. Two issues are occuring:

1. No lakeh and rsfil files are generated to include in the warm run as inputs.
2. The dis.nc produces discharges much larger for the same dates and places as HERA, while in theory they should be the same. What am I doing wrong?

My goal is to make sure that my setup is correct in order to use lisflood for accurate discharge predictions.

[StefaniaGrimaldi]

Dear @usereight8,

we are pleased to know about your interest in OS LISFLOOD and CEMS EFAS.

Your messages points to different issues. I suggest proceeding step by step.

I understand that you aim to replicate CEMS EFAS reanalysis. Please be aware that CEMS EFAS reanalysis is available from https://ewds.climate.copernicus.eu/datasets/efas-historical?tab=overview .
The static maps are available from https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/LISFLOOD_static_and_parameter_maps_for_EFAS/
The meteorological forcings can be downloaded from https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/meteorological_forcings/EMO-1arcmin/
Information on EFAS calibration and reanalysis can be found at https://confluence.ecmwf.int/display/CEMS/EFAS+v5.0

HERA is a different reanalysis dataset, which is described in https://essd.copernicus.org/articles/17/293/2025/essd-17-293-2025.pdf

OS LISFLOOD runs: to replicate the results, you will need to use exactly the same settings, parameters, meteo, and static input maps. This file might be useful for such purpose https://github.com/ec-jrc/lisflood-calibration/blob/master/integration/settings_EFAS5.xml

I was wondering whether you already performed a prerun (if so, for which period).

Are you modelling the extended European domain or a subset? If your modelling domain does not have lakes/reservoirs, nore lakeh.nc neither resfil.nc are written in the output folder.

I hope that this initial answer helps,
we are available for further exchanges,
on behalf of OS LISFLOOD developers team,
kind regards,
Stefania

[usereight8]

Dear @StefaniaGrimaldi,

Thank you very much for your response. My end goal is to be able to apply EVA to future predictions of lisflood for a subset of Europe to attain a return period for each point of this subset. I though of using the HERA dataset's discharges first to fit the GEV distribution.

However to make sure first that my lisflood predictions are accurate I thought of conducting a test for the whole of Europe, i.e. to use the static maps and parameters from here: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/LISFLOOD_static_and_parameter_maps_for_EFAS/, the climate inputs from here: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/HERA/VER1-0/Data/NetCDF/climate_inputs/ hoping that I get in return something close to the HERA discharges: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/HERA/VER1-0/Data/NetCDF/river_discharge/.

I ran a 1-year prerun (with InitLisflood=1) for 2016 (https://github.com/usereight8/lisflood/blob/main/cold.xml) and then I ran a warm run for a couple of days in 2017. However when I compared the dis.nc output of the warm run against the discharges of the HERA dataset for 2017 for a single point the former were 3 times larger. Also lakes and reservoirs outputs were missing so I deactivated them in the warm run.

Is the logic behind this test incorrect?

From what I understand from your response I should use LISVAP to produce evapotranspirations from EMO-1: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/meteorological_forcings/EMO-1arcmin/, run lisflood using the template you suggested: https://github.com/ec-jrc/lisflood-calibration/blob/master/integration/settings_EFAS5.xml, use the same static parameters as above: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/LISFLOOD_static_and_parameter_maps_for_EFAS/ and then I should get a result close to the EFAS reanalysis which I can download here:https://ewds.climate.copernicus.eu/datasets/efas-historical?tab=overview.

Is this what you are suggesting instead?

[StefaniaGrimaldi]

Dear @usereight8,

thanks for proving further details on your initial tests and final aim.

OF LISFLOOD behavior in your initial tests is as expected.
The prerun (or initialization run, InitLisflood=1) does not produce end files for lakes and reservoirs. The prerun is designed to deliver lzavin, avgids, and end files of soil state variables. Please be aware that 1 year of prerun is not enough. The prerun should explore most of the climatic behaviour of the run. You can find more information in the OS LISFLOOD user guide: https://ec-jrc.github.io/lisflood-code/3_step5_model-initialisation/
The correct sequence is: prerun, coldstart, warmstart (if needed).
The large values on the initial days of the prerun are expected and they are due to the initial conditions in the channels (half of bankful). The results of the prerun cannot be compared to renalaysis data.

We recommend users that approach OS LISFLOOD for the first time to read the documentation and user guide in full, and then perform tests using pre-compiled set-ups provided in https://github.com/ec-jrc/lisflood-usecases, https://github.com/ec-jrc/lisflood-usecases/tree/mekong_usecase/LF_mekong_usecase

OS LISFLOD has many settings and the definition of the correct simulations pipeline is not trivial on a first attempt, therefore, I recommend experimenting with the more simple set-up of the OS LISFLOOD use cases, before attempting to replicate EFAS or HERA reanalysis.

I pointed you to the EMO dataset and to the EWDS for the meteorological inputs and the reanalysis data, respectively, because the title of the issue is "EFAS replication". You can surely decide whether to replicate EFAS or HERA. In both cases, your set-up must be consistent to the set-up of the original simulation (e.g. HERA makes use of different land cover and reservoirs input data every year).

Finally, I would like to share some information on EFAS flood thresholds, which might be useful for your work. EFAS flood thresholds are computed using the following OS LISFLOOD utility: https://github.com/ec-jrc/lisflood-utilities/tree/master?tab=readme-ov-file#thresholds . EFAS thresholds for each pixel of the computational domain can be downloaded from https://confluence.ecmwf.int/display/CEMS/Auxiliary+Data#AuxiliaryData-EFASAuxiliaryData "EFAS Flood Thresholds".

I hope that this answer helps,
we are available for further exchanges,
on behalf of OS LISFLOOD developers team,
kind regards,
Stefania

[usereight8]

Dear @StefaniaGrimaldi,

Your response was super helpful coupled with the crystal clear tutorial: https://github.com/ec-jrc/lisflood-usecases/tree/mekong_usecase/LF_mekong_usecase. It has completely cleared the confusion for me regarding initialization, cold and warm runs.

If you would be so kind I have two last questions.

1. Regarding CalendarDayStart I can see that:

Initialization
<textvar name="CalendarDayStart" value="02-01-1979 00:00"/> <textvar name="StepStart" value="02-01-1979 00:00"/> <textvar name="StepEnd" value="01-01-2020 00:00"/>

Cold run
<textvar name="CalendarDayStart" value="02-01-1979 00:00"/> cold <textvar name="StepStart" value="02-01-1979 00:00"/> <textvar name="StepEnd" value="01-01-1990 00:00"/>

Warm run
<textvar name="CalendarDayStart" value="02/01/1979 00:00"/> warm <textvar name="StepStart" value="02/01/1990 00:00"/> <textvar name="StepEnd" value="01/01/2000 00:00"/>

Should CalendarDayStart always be equal to the datetime that the meteorological forcings begin that must be the same with the datetime that the dynamic maps begin? So in the case of EFAS, if the dynamic maps begin at 01/01/1990 00:00 then I must always use meteorological netcdf files that begin also at 01/01/1990 00:00 and then CalendarDayStart should be 01/01/1990 06:00?

2. I have noticed that initialization doesn't run without laiofday.txt however I can find no instructions on how to create this file which doesn't exist here: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/.

How can I create it? Do I need other tables like firstdayofmonth.txt apart those already uploaded: https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/CEMS-EFAS/LISFLOOD_static_and_parameter_maps_for_EFAS/Lakes_resersoirs_ID_tables/?

[StefaniaGrimaldi]

Dear @usereight8,

we are pleased to know that the previous answers were useful.

We recommend setting the CalendarDayStart equal to the first step of the meteorological forcings (please note the convention https://ec-jrc.github.io/lisflood-code/2_ESSENTIAL_time-management/). The CalendarDayStart is used to write the timestamp of the .tss files. I

Neither laiofday.txt nor firstdayofmonth.txt are required by the current version of OS LISFLOOD (master branch: https://github.com/ec-jrc/lisflood-code). All the maps and tables required by EFAS set-up (EFAS v5.x, https://confluence.ecmwf.int/display/CEMS/EFAS+v5.0) are avialble from the online repository.

I hope that this answer clarified your doubts,
we are available for further exchanges,
on behalf of OS LISFLOOD developers team,
kind regards,
Stefania

[usereight8]

Dear @StefaniaGrimaldi ,

Thank you very much for your very helpful responses.

I am trying to implement your suggestions.

I will get back to you with another ticket in case I face another difficulty.